A comparison of ML, WLSMV and Bayesian methods for multilevel structural equation models in small samples: A simulation study

Research output: Journal contributionsJournal articlesResearchpeer-review

Standard

A comparison of ML, WLSMV and Bayesian methods for multilevel structural equation models in small samples: A simulation study. / Holtmann, Jana; Koch, Tobias; Lochner, Katharina et al.
In: Multivariate Behavioral Research, Vol. 51, No. 5, 02.09.2016, p. 661-680.

Research output: Journal contributionsJournal articlesResearchpeer-review

Harvard

APA

Vancouver

Bibtex

@article{838197da39304d64861dc158dc6296c0,
title = "A comparison of ML, WLSMV and Bayesian methods for multilevel structural equation models in small samples: A simulation study",
abstract = "Multilevel structural equation models are increasingly applied in psychological research. With increasing model complexity, estimation becomes computationally demanding, and small sample sizes pose further challenges on estimation methods relying on asymptotic theory. Recent developments of Bayesian estimation techniques may help to overcome the shortcomings of classical estimation techniques. The use of potentially inaccurate prior information may, however, have detrimental effects, especially in small samples. The present Monte Carlo simulation study compares the statistical performance of classical estimation techniques with Bayesian estimation using different prior specifications for a two-level SEM with either continuous or ordinal indicators. Using two software programs (Mplus and Stan), differential effects of between- and within-level sample sizes on estimation accuracy were investigated. Moreover, it was tested to which extent inaccurate priors may have detrimental effects on parameter estimates in categorical indicator models. For continuous indicators, Bayesian estimation did not show performance advantages over ML. For categorical indicators, Bayesian estimation outperformed WLSMV solely in case of strongly informative accurate priors. Weakly informative inaccurate priors did not deteriorate performance of the Bayesian approach, while strong informative inaccurate priors led to severely biased estimates even with large sample sizes. With diffuse priors, Stan yielded better results than Mplus in terms of parameter estimates.",
keywords = "Social Work and Social Pedagogics, multilevel structural equation modeling, multilevel item response theory, Bayesian statistics, sample size, Monte Carlo simulation",
author = "Jana Holtmann and Tobias Koch and Katharina Lochner and Michael Eid",
year = "2016",
month = sep,
day = "2",
doi = "10.1080/00273171.2016.1208074",
language = "English",
volume = "51",
pages = "661--680",
journal = "Multivariate Behavioral Research",
issn = "0027-3171",
publisher = "Psychology Press Ltd",
number = "5",

}

RIS

TY - JOUR

T1 - A comparison of ML, WLSMV and Bayesian methods for multilevel structural equation models in small samples: A simulation study

AU - Holtmann, Jana

AU - Koch, Tobias

AU - Lochner, Katharina

AU - Eid, Michael

PY - 2016/9/2

Y1 - 2016/9/2

N2 - Multilevel structural equation models are increasingly applied in psychological research. With increasing model complexity, estimation becomes computationally demanding, and small sample sizes pose further challenges on estimation methods relying on asymptotic theory. Recent developments of Bayesian estimation techniques may help to overcome the shortcomings of classical estimation techniques. The use of potentially inaccurate prior information may, however, have detrimental effects, especially in small samples. The present Monte Carlo simulation study compares the statistical performance of classical estimation techniques with Bayesian estimation using different prior specifications for a two-level SEM with either continuous or ordinal indicators. Using two software programs (Mplus and Stan), differential effects of between- and within-level sample sizes on estimation accuracy were investigated. Moreover, it was tested to which extent inaccurate priors may have detrimental effects on parameter estimates in categorical indicator models. For continuous indicators, Bayesian estimation did not show performance advantages over ML. For categorical indicators, Bayesian estimation outperformed WLSMV solely in case of strongly informative accurate priors. Weakly informative inaccurate priors did not deteriorate performance of the Bayesian approach, while strong informative inaccurate priors led to severely biased estimates even with large sample sizes. With diffuse priors, Stan yielded better results than Mplus in terms of parameter estimates.

AB - Multilevel structural equation models are increasingly applied in psychological research. With increasing model complexity, estimation becomes computationally demanding, and small sample sizes pose further challenges on estimation methods relying on asymptotic theory. Recent developments of Bayesian estimation techniques may help to overcome the shortcomings of classical estimation techniques. The use of potentially inaccurate prior information may, however, have detrimental effects, especially in small samples. The present Monte Carlo simulation study compares the statistical performance of classical estimation techniques with Bayesian estimation using different prior specifications for a two-level SEM with either continuous or ordinal indicators. Using two software programs (Mplus and Stan), differential effects of between- and within-level sample sizes on estimation accuracy were investigated. Moreover, it was tested to which extent inaccurate priors may have detrimental effects on parameter estimates in categorical indicator models. For continuous indicators, Bayesian estimation did not show performance advantages over ML. For categorical indicators, Bayesian estimation outperformed WLSMV solely in case of strongly informative accurate priors. Weakly informative inaccurate priors did not deteriorate performance of the Bayesian approach, while strong informative inaccurate priors led to severely biased estimates even with large sample sizes. With diffuse priors, Stan yielded better results than Mplus in terms of parameter estimates.

KW - Social Work and Social Pedagogics

KW - multilevel structural equation modeling

KW - multilevel item response theory

KW - Bayesian statistics

KW - sample size

KW - Monte Carlo simulation

UR - http://www.scopus.com/inward/record.url?scp=84984906875&partnerID=8YFLogxK

U2 - 10.1080/00273171.2016.1208074

DO - 10.1080/00273171.2016.1208074

M3 - Journal articles

C2 - 27594086

VL - 51

SP - 661

EP - 680

JO - Multivariate Behavioral Research

JF - Multivariate Behavioral Research

SN - 0027-3171

IS - 5

ER -

Recently viewed

Publications

  1. Image compression based on periodic principal components
  2. Using CNNs to Detect Graphical Representations of Structural Equation Models in IS Papers
  3. Exploring the limits of graph invariant- and spectrum-based discrimination of (sub)structures.
  4. Data based analysis of order processing strategies to support the positioning between conflicting economic and logistic objectives
  5. Linear free vibrations with uncertain initial conditions
  6. Appendix A: Design, implementation, and analysis of the iGOES project
  7. Digging into the roots
  8. Dividing Apples and Pears: Towards a Taxonomy for Agile Transformation
  9. Optimization of 3D laser scanning speed by use of combined variable step
  10. Gain Adaptation in Sliding Mode Control Using Model Predictive Control and Disturbance Compensation with Application to Actuators
  11. Overcoming Multi-legacy Application Challenges through Building Dynamic Capabilities for Low-Code Adoption
  12. Modelling biodegradability based on OECD 301D data for the design of mineralising ionic liquids
  13. Effectiveness of a Web-Based Cognitive Behavioural Intervention for Subthreshold Depression
  14. Switching between reading tasks leads to phase-transitions in reading times in L1 and L2 readers
  15. Challenge-oriented policy making and innovation systems theory: reconsidering systemic instruments
  16. Cross-case knowledge transfer in transformative research: enabling learning in and across sustainability-oriented labs through case reporting
  17. Towards Advanced Learning in Dispatching Rule-Based Scheuling
  18. Kit based motion generator for a soft walking robot